It’s been a hectic month in the upper echelons of open-weight AI, with flagship models released by Qwen, MiniMax, Z.ai, Ant Ling, and StepFun. All eyes remain fixed on the imminent release of DeepSeek V4, as rumors continue to build momentum around it. Beyond the major frontier models, this problem is less severe across the long tail of specialized modalities and smaller model scales. We’re monitoring all these new releases using our latest Relative Adoption Metrics (RAM), which normalizes download counts against comparable models within the same size category. This has already proven to be an incredibly valuable tool for us, shining a light on underrated models like GPT-OSS — which is off the charts in terms of downloads and stands as the most popular American open-weights model since Llama 3.1. A RAM score above 1 indicates that the model is on pace to become one of the top 10 most-downloaded models of all time in its size category. We’re especially curious to see how early adoption of the smaller Qwen 3.5 dense models compares to Qwen 3 — striking a balance between Qwen’s rapidly expanding brand and a more complex hybrid architecture that challenges the boundaries of certain open-source tools. Below is a summary of the RAM scores for some popular models released in late 2023, which clearly positions Kimi K2 Thinking and several OCR models as the top performers. DeepSeek V3.2 and the company’s other recent large models have dramatically underperformed compared to DeepSeek’s earlier 2025 releases. (Time shown is days since release.) Qwen23-210B-A22026B by Qwen: The long-awaited update to the Qwen series has finally arrived.
Interconnects AI